AUV Tri-TON is a hovering type autonomous underwater vehicle developed by the University of Tokyo, launched in 2011. The vehicle was constructed as a testbed under the governmental project to develop instruments to estimate ore reserves in underwater hydrothermal deposits. The vehicle's mission is to obtain dense, large-area 3D image of hydrothermal vent fields, in collaboration with a seafloor station. The information will be also used for environmental assessments, mine planning, and educational activities.Although the vehicle is not equipped with an inertial navigation system (INS), the vehicle can estimate its position in real-time with a precision enough for rough photo-mosaicking, owing to the mutual acoustic positioning with the station. The vehicle has two suites of imaging instruments looking forward and downward directions in order to image whole surface of bumpy hydrothermal vent fields. The vehicle has been tested through a series of experiments at tanks and real fields. In April 2012 the vehicle was deployed to the hydrothermal vent field of Kagoshima Bay in Japan and succeeded in observing seafloor with the area of around 200 square meters.
Although the vast amount of information collected by AUVs brings significant benefit to oceanographic research, it is necessary to develop methods to analyze the large volumes of data, in order to avoid accumulation of unused information. Automatic data processing and analysis are key technologies necessary to cope with this problem. We propose a robust, automated method for detection and volume determination of tubeworm colonies using visual and geometric features obtained during underwater robotic surveys, on the condition that the position of the sensors are provided. The tubeworm is a characteristic benthos of hydrothermal vent fields. The proposed method achieves robustness against sensor noise by using both geometric and visual features for identification. First, the tubeworm candidates are obtained as a three-dimensional region between the measured bathymetry of the region and an estimation of the seafloor topology without tubeworms. Next, the tubeworm candidates are verified through frequency analysis of corresponding images. The performance of this method was verified using a data set obtained by the AUV Tri-Dog 1 at Tagiri vent field, Kagoshima bay in Japan.
When robots work in a cluttered environment, the constraints for motions change frequently and the required action can change even for the same task. However, planning complex motions from direct calculation has the risk of resulting in poor performance local optima. In addition, machine learning approaches often require relearning for novel situations. In this paper, we propose a method of searching appropriate motions by using conditional Generative Adversarial Networks (cGANs), which can generate motions based on the conditions by mimicking training datasets. By training cGANs with various motions for a task, its latent space is fulfilled with the valid motions for the task. The appropriate motions can be found efficiently by searching the latent space of the trained cGANs instead of the motion space, while avoiding poor local optima. We demonstrate that the proposed method successfully works for an object-throwing task to given target positions in both numerical simulation and real-robot experiments. The proposed method resulted in three times higher accuracy with 2.5 times faster calculation time than searching the action space directly.
A B S T R A C TAutonomous underwater vehicles (AUVs) can operate without the need for human control or tether cables as long as there is sufficient energy. AUVs have recently been used for seafloor imaging. Visual observation by AUVs provides high-resolution color information of the seafloor. However, conventional observation techniques that follow a prespecified path offer limited coverage because it is impossible for operators to build a suitable path in unknown rough terrain. A flawed prespecified path will produce incomplete observation. If unobserved areas are found during postprocessing, another dive is necessary, which increases the total cost. To overcome this problem, the authors have proposed a path replanning method to realize high-coverage observation in one dive. With this method, the AUV evaluates unobserved areas after the first prespecified observation; if unobserved areas are found, the AUV recreates an appropriate path to cover what was missed. The validity of the proposed method was previously evaluated using an artificial target in a tank and in shallow seas at a depth of approximately 35 m. In this study, the feasibility of the method was validated in a more challenging setting: experimental data were taken from a hydrothermal vent field in Kagoshima Bay, Japan.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.